Skip to main content

Table 4 External validation: segmentation performance on images with artifacts

From: Cardiovascular magnetic resonance images with susceptibility artifacts: artificial intelligence with spatial-attention for ventricular volumes and mass assessment

 

LV ED

LV ES

DSC

HD (mm)

Recall

Precision

DSC

HD (mm)

Recall

Precision

CNN vs GT

0.94 (0.91–0.96)

5.2 (3.2–8.5)

0.94 (0.88–0.97)

0.96 (0.94–0.97)

0.92 (0.94–0.94)

4.8 (3.2–8.5)

0.92 (0.82–0.95)

0.94 (0.91–0.97)

 

RV ED

RV ES

DSC

HD (mm)

Recall

Precision

DSC

HD (mm)

Recall

Precision

CNN vs GT

0.90 (0.84–0.92)

8.2 (5.8–12.9)

0.87 (0.80–0.93)

0.92 (0.89–0.94)

0.88 (0.80–0.91)

8.2 (4.8–13.1)

0.85 (0.77–0.90)

0.92 (0.88–0.94)

 

LVM ED

LVM ES

DSC

HD (mm)

Recall

Precision

DSC

HD (mm)

Recall

Precision

CNN vs GT

0.79 (0.75–0.83)

5.0 (3.2–6.5)

0.87 (0.81–0.91)

0.75 (0.68–0.79)

0.83 (0.78–0.87)

4.4 (3.5–6.8)

0.88 (0.83–0.92)

0.79 (0.71–0.85)

  1. Values are reported as median (interquartile range)
  2. Segmentation performance on images with artifacts for the proposed CNN compared to the manual gold standard (GT)
  3. Abbreviations as in Table 2